Received June 9, 2020, accepted July 4, 2020, date of publication July 7, 2020, date of current version July 20, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.3007786

Spatiotemporal Change and Landscape PatternVariation of Eco-Environmental Quality inJing-Jin-Ji Urban AgglomerationFrom 2001 to 2015JIANWAN JI 1,2, SHIXIN WANG 1, YI ZHOU 1,WENLIANG LIU 1, AND LITAO WANG 11Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, China2University of Chinese Academy of Sciences, Beijing 100049, China

Corresponding authors: Shixin Wang (wangsx@radi.ac.cn) and Wenliang Liu (liuwl@radi.ac.cn)

This work was supported by the National Key Research and Development Program of China under Grant 2017YFB0503805.

ABSTRACT Identifying the changes and relationships between regional eco-environment quality and land-scape pattern in an urban agglomeration have a great significance in realizing sustainable development goal.However, limited research has been performed to understand the spatiotemporal change of eco-environmentquality, the variation of landscape pattern, and their relationship in an urban agglomeration. This studyselected the Jing-Jin-Ji (JJJ) urban agglomeration as the study area. A comprehensive index, the remotesensing ecological index (RSEI), was utilized to understand the eco-environment spatiotemporal changeand landscape pattern variation at class-level and landscape-level of JJJ during 2001∼2015, then, theirrelationship was explored. The major conclusions were as follows: (1) The average RSEI value of JJJincreased from 0.43 to 0.46, which represented that the eco-environment of JJJ had improved in the fourteenyears. Among it, the improved region was mainly located in Zhangjiakou city, while the degraded region wasmainly distributed in the eastern Hebei plain. (2) The landscape characteristics of entire JJJ eco-environmentwere becoming more aggregated, connected, diverse, and regular. However, fair, moderate, and good gradeswere getting more concentrated and continuous; poor grade indicated a more fragmented and disconnectedtrend; excellent grade displayed an expanded and concentrated situation. (3) Human factors have anincreasing influence on regional eco-environment changes. (4) Fair, moderate, and good grades showeda more dominant and stronger influence on the variation of landscape pattern in JJJ. Specifically, the fairgrade had a positive correlation with the variation of landscape pattern, while moderate and good gradeshad a negative one. All of these conclusions could be valuable information for relevant decision-makers inmanaging or achieving the optimal eco-environment landscape pattern.

INDEX TERMS Jing-Jin-Ji urban agglomeration, landscape pattern, remote sensing ecological index,spatiotemporal change.

I. INTRODUCTIONThe eco-environment, defined as ‘‘the total quantity and qual-ity of water resources, land resources, biological resourcesand climate resources that affect human survival and devel-opment’’, is a social-economic-natural compound system.It not only provides human natural resources and livingenvironment service but also is the foundation and core of

The associate editor coordinating the review of this manuscript and

approving it for publication was Weimin Huang .

regional social and economic sustainable development [1].Its quality can effectively indicate the degree of coordi-nation between human production activities and the envi-ronment of a region [2]. Since the implementation of thepolicy of reform and opening up in 1978, great changes havetaken place in mainland China, especially in the aspects ofspatial urbanization, population expansion, industrialization,etc. [3], [4]. However, accompanied by the spatial urban-ization process, the eco-environment has also been greatlyinfluenced, for example, water and soil loss [5], urban heat

125534 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ VOLUME 8, 2020

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

islands [6], a decrease of vegetation coverage [7], air pollu-tion [8], and so forth, which can pose a threat to the realiza-tion of sustainable development goals. Hence, it is of greatimportance to assess and analyze regional eco-environmentquantitatively.

Since the concept of Ekistics was promoted by Dox-iadis in the 1950s, developed countries had carried outsome researches to analyze and evaluate urban livingenvironment quality [9], [10]. In 1992, Fu evaluatedeco-environmental qualities of China systematically [11].To date, numerous studies have been conducted to assessregional eco-environment at various scales, such as inter-national [12], intercontinental [13], national [14], provin-cial [15], city [16], etc. As for the evaluation methods,it can be mainly divided into two types, which are qual-itative and quantitative evaluation methods [1]. Till now,there has developed several methods, for example, indexevaluation method [17], analytic hierarchy process [18],ecological footprint method [19], artificial neural net-work evaluation method [20], [21], matter element analy-sis method [22], fuzzy integrated assessment method [23],and so forth [1], [24]. Generally speaking, compared withqualitative evaluation, quantitative evaluation gives moreobjective judgement about the eco-environment. However,numerous traditional methods could only evaluate thewhole regional eco-environment by one calculated value,which is unable to provide the evaluation value of anylocation.

Advances in remote sensing technology have showngreat potential in evaluating numerous aspects of regionaleco-environment at various scales, which owes to the prin-cipal that remote sensing imagery acquires reflected radia-tion of the earth surface. For example, regional vegetationsituation can be detected by one widely used index—Normalized Difference Vegetation Index (NDVI) [25]; EVI-derived ecosystem functional attributes (EFAs), promotedby Alcaraz-Segura in 2017 [26], could be applied as oneimportant biodiversity variables in species distribution mod-els. Besides, land surface temperature (LST), acquired fromremote sensing thermal imagery has also been an importantparameter to analyze the urban heat island and the dynamicsand evolution of regional thermal environments [27]. Over-all, a single remote-sensing index can only reflect a limitedaspect. Considering the complexity of an eco-environment,aggregated remote sensing index has drawn the attentionof scholars all over the world. He et al. [28] developed acomprehensive evaluation index (CEI) to assess urban envi-ronment change in China; Wei et al. [29] integrated sixindexes to evaluate the environment. What’s more, numer-ous researchers combined remote sensing data and otherdatasets to assess regional eco-environment. Wei et al. [30]integrated 23 indices to assess environmental vulnerability;Chang et al. [31] constructed an index system combined14 indices to evaluate the ecological environment; Chai andLha [32] assessed ecological environmental quality (EEQ)by selecting key indicators; Sun et al. [33] evaluated the

eco-environmental quality of Hainan island by establishingan eco-environmental quality index (EQI), which was devel-oped and published by the ministry of ecology and environ-ment of China.

Generally, single remote sensing indices or existing aggre-gated remote sensing indices mostly assess one certainaspect of a regional eco-environment. Constructing one indexsystem is an effective way to comprehensively evaluate aregional eco-environment, however, index system mostlyrequires multi-source datasets, which is time-consumingand inconvenient, moreover, the establishment of an appro-priate sub-index system can exert a great influence onthe final evaluation result [34]. So, does there exist oneaggregated remote sensing index that can comprehensivelyevaluate a regional eco-environment? One index promotedby Prof. Xu in 2013 has made some progress [35]. Thisindex is named as remote sensing ecological index (RSEI),it encompasses four sub-indexes representing climatic andland-surface biophysical variables [36]. To be specific, thesefour sub-indexes are normalized difference vegetation index(NDVI), wetness (WET), normalized difference build-up andsoil index (NDBSI) and land surface temperature (LST),respectively. Among it, NDVI represents the greenness aspectof a regional eco-environment; WET represents the wetnessaspect; NDBSI represents the dryness aspect; LST representsthe heat aspect. Spatial principal component analysis (SPCA)is an effective way to aggregate the most valuable infor-mation [37]. By utilizing the SPCA method, RSEI can beacquired. Till now, it has been widely applied in numerousstudies, which has proven to be an effective and convenientindex in quickly evaluating a regional eco-environment, suchas Fuzhou city [38]–[40]; Xiong’an New Area [41]; Nan-chang city [42]; Dingcheng district in Changde city [43];Zhengzhou city [44], etc. However, these existing studiesmostly applied RSEI to evaluate regional eco-environment atcity level based onmedium resolution remote sensing images,like Landsat series image [38], which failed to evaluate atthe large region level. The major reason was that at thelarge region level, it was inappropriate and difficult to applythis index. For instance, one large region mostly requiresmultiple Landsat images to cover the entire region, however,due to the cloud pollution and the 16 days revisit period,it was extremely hard to acquire all cloud-free images atthe same time. To solve this problem, integrating this indexwith MODIS datasets and Google Earth Engine platform hasshown great potential.

Landscape ecology is one subject that aims to study andimprove the relationships between specific ecosystems andtheir ecological processes [45]. Landscape pattern, as one ofthe key research topics of landscape ecology [46], focuseson the quantification of changes in the land elements’configurations and compositions based on landscape met-rics [47]–[49]. Moreover, landscape pattern is consideredto be an important indicator of landscape heterogeneityand its effects on a variety of ecological processes [50].Monitoring land use/cover changes and landscape pattern

VOLUME 8, 2020 125535

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

analysis have drawn much concern from scholars around theworld [51]. It was concluded that land use/cover changesand landscape pattern variations have certain ecologicaleffects [52]. A literature review revealed that three mainissues had been studied, which were the changes in land-scape patterns [53], [54], the influences of these changes onecological processes [55], [56], and the relationship betweenlandscape pattern changes and their driving forces [57]–[59].For example, Wang et al. [47] found that cultivated land andwater bodies had a close relationship with four landscape-level metrics. Besides these studies, the spatial configurationof the urban thermal environment and its relationship withlandscape pattern metrics have been investigated by someresearchers [45], [60]. To date, however, existing studiesseldom investigated the relationship between different eco-environment grades and landscape metrics.

An urban agglomeration is a highly developed spatialform of integrated cities, with a research history traced backto 100 years ago [61]. In 2014, the Chinese governmentissued a development roadmap of ‘‘National New Urban-ization Plan’’, which clearly pointed out to optimize andupgrade eastern urban agglomerations, establish a coordi-nated mechanism for urban agglomerations development,and to achieve the green development goals based on pro-tecting the eco-environment [62]. Till now, China has pro-posed building a hierarchical urban agglomeration systemwith five national-level large urban agglomerations, nineregional-level medium-sized urban agglomerations and sixsub-regional-level small-sized urban agglomerations [61].Jing-Jin-Ji urban agglomeration (JJJ), which is one of thefive large urban agglomerations and considered as ‘‘the threeengines of China’s economic growth’’ in the 21st century(the other two are Yangtze Delta urban agglomeration andPearl River Delta urban agglomeration), was selected as thestudy area. The aims of this study were to (1) analyze thespatiotemporal changes of RSEI; (2) evaluate the landscapepattern variation of RSEI grades based on class-level andlandscape-level metrics; (3) identify the relationship betweenRSEI grades and landscape pattern.

II. MATERIALS AND METHODSA. STUDY AREAJJJ is located in north China (36◦05′∼42◦40′N,113◦27′∼119◦50′E) and covers approximately 218,000 km2.It includes two municipalities and eleven prefecture-levelcities, which are Beijing (BJ), Tianjin (TJ), Shijiazhuang(SJZ), Tangshan (TS), Qinhuangdao (QHD), Handan (HD),Xingtai (XT), Baoding (BD), Zhangjiakou (ZJK), Chengde(CD), Cangzhou (CZ), Langfang (LF) and Hengshui (HS)(Figure 1). This region has a temperate semi-humid andsemi-arid monsoon climate with the average temperature inJuly and annual precipitation of 18∼27 ◦C and 524.4 mmrespectively. In 2017, the total population and gross domes-tic product of JJJ reached 95.74 million and 8058.04 bil-lion yuan, which accounted for 6.89% and 9.77% of thatof the whole country (http://www.stats.gov.cn). However,

FIGURE 1. Location of the study area.

along with the fast urbanization, quantitatively evaluatingthe change of JJJ’s eco-environment has become a topic ofdiscussion.

B. DATA PREPARATION AND RSEI CONSTRUCTIONIn this paper, MOD09A1 and MOD11A2 datasets were uti-lized to construct RSEI. Specifically, MOD09A1 dataset pro-vides an estimate of the 8-day Terra MODIS seven bandssurface spectral reflectance corrected for atmospheric con-ditions such as aerosols, gasses, and Rayleigh scattering at500 m resolution; MOD11A2 dataset provides an average8-day land surface temperature (LST) at 1000 m resolu-tion based on the generalized split-window algorithm [63].In order to keep the results comparable, both datasets inthe time span of 1st June to 31st October were processedon Google Earth Engine (GEE) platform (https://code.earthengine.google.com).

RSEI is composed of four sub-indexes, which are normal-ized difference vegetation index (NDVI), wetness (WET),normalized difference build-up and soil index (NDBSI) andland surface temperature (LST). Different from previousstudies, RSEI integrates MODIS high-temporal datasets,making it possible to assess large-scale regional eco-environment. Except that LST is directly acquired fromMOD11A2 dataset, NDVI [25], WET [64] and NDBSI [35]are acquired based on the following formulas, where ρ is the

125536 VOLUME 8, 2020

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

band surface reflectance, blue, green, red, nir, mir1, mir2,mir3 are the MODIS bands at 459-479 nm, 545-565 nm,620-670 nm, 841-876 nm, 1230-1250 nm, 1628-1652 nm,and 2105-2155 nm respectively.

NDVI = (ρnir − ρred ) / (ρnir + ρred ) (1)

WET = 0.1084× ρred + 0.0912× ρnir+ 0.5065× ρblue + 0.4040× ρgreen− 0.2410× ρmir1 − 0.0.4658× ρmir2− 0.5306× ρmir3 (2)

NDBSI =12

2×ρmir2ρmir2+ρnir

−ρnir

ρnir+ρred−

ρgreenρgreen+ρmir2

2×ρmir2ρmir2+ρnir

+ρnir

ρnir+ρred+

ρgreenρgreen+ρmir2

+12

{(ρmir2 + ρred )− (ρmir2 + ρblue)

(ρmir2 + ρred )+ (ρmir2 + ρblue)

}(3)

After acquired all 8-day sub-indexes results during theperiod, a final average value of each sub-index was firstlycalculated andwas rescaled to 0∼1, then principal componentanalysis (PCA) was performed in ArcGIS 10.6 software.Normally, the first component of PCA (PC1) integrates thelargest information of the input dataset. Therefore, PC1 wasadopted to derive original RSEI value and the expressioncould be written as follows.

RSEIorigin = 1− PC1 [f (NDVI ,WET ,NDBSI ,LST )] (4)

Finally, RSEI was obtained by rescaling to 0∼1. The for-mula of rescaling was as follows.

Xrescale = (Xi − Xmin) / (Xmax − Xmin) (5)

where Xrescale represents the rescaled result; Xmax and Xminrepresent the maximum and minimum value of X ; Xi meansX value at the ith pixel. Figure 2 is a flowchart.

C. RSEI SPATIOTEMPORAL CHANGE ANALYSISNormally, spatiotemporal change includes two aspects, whichare spatial change and temporal change. Before analyze RSEIspatiotemporal characteristics, we set 0.2 as the interval todivide RSEI into five grades, which are poor [0.0, 0.2), fair[0.2, 0.4), moderate [0.4, 0.6), good [0.6, 0.8), and excellent[0.8, 1.0], respectively. The dynamic model is one of thefrequently used models [47]. It can reflect the change degreeof a certain RSEI grade in a certain period quantitively. Theformula is as follows.

D =Tend − Tstart

Tstart×

1P× 100% (6)

where D is the dynamic degree; Tstart and Tend indicate thearea of a certain RSEI grade at the start and end of thecomparison period, respectively; P is the time interval.Besides, the transition matrix method is applied to deeply

understand the area transition situation between different

FIGURE 2. RSEI construction flowchart.

RSEI grade [65]. The formula is as follows.

S =

S11 S12 · · · S1nS21 S22 · · · S2n...

.... . .

...

Sn1 Sn2 · · · Snn

(7)

VOLUME 8, 2020 125537

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

TABLE 1. The brief introduction of selected landscape metrics.

where S represents the area of a certain RSEI grade; n isthe number of RSEI grade; Sij indicates the transformed areafrom grade i to grade j in a certain period; Sii represents theunchanged area in a certain period.

D. LANDSCAPE METRICS SELECTION AND CALCULATIONNumerous landscape metrics have been promoted to explainthe relationships between ecological processes and spatialpatterns. Till now, many programs have been developed andpromulgated, which acquires numerous landscape metricsbecome more accessible [47]. Among it, Fragstats 4.2 soft-ware is one of the most used platforms, it can compute hun-dreds of landscape metrics for three levels: patch level, classlevel, and landscape level [66]. However, previous studieshave found that redundancy between various metrics waswidely existed [51], [67].

Therefore, the selection of an optimal landscape metricsis the key step for further landscape pattern analyses. Here,correlation analysis of 54 metrics, including 24 metrics at theclass level and 30 metrics at the landscape level, was firstlyperformed, then criterion |r| ≥ 0.9 was applied to excludeunsatisfied metrics [67]; Next, combined with previous stud-ies [47], [51], [60], [68], [69], ten represented landscape met-rics were finally selected to quantitatively evaluate landscapepattern variations at the different period with the support ofFragstats 4.2 (use 8 cell neighbour rule). Table 1 is the briefintroduction of ten selected landscape metrics.

III. RESULTSA. RSEI SPATIOTEMPORAL CHANGE ANALYSISFigure 3 illustrates the spatial distribution of four years ofRSEI. The average RSEI value of 2001, 2005, 2010 and2015 were 0.43, 0.42, 0.51 and 0.46 respectively, which

125538 VOLUME 8, 2020

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

FIGURE 3. Spatial distribution of RSEI in JJJ.

presented an inverted ‘‘N’’ change trend. Generally, the eco-environment of JJJ degraded slightly (2001∼2005), thenimproved sharply (2005∼2010), and deteriorated again in thelast period (2010∼2015); however, from the perspective ofthe entire period, the eco-environment of JJJ had improved.To be specific, it could be found that the northern partof JJJ had a better eco-environment compared with north-western, central, and south-eastern part, which was followingthe actual situation, as the northern part was mainly coveredwith vegetation and belonged to a mountainous area namedYanshan Mountain.

Table 2 displays RSEI five grades statistical result. It couldbe found that area in poor eco-environment contracted from2001 to 2010, and slightly expanded towards 2015; The fairgrade expanded and contracted alternately, with the greatestshrinkage from 2005∼2010 (−28.71%); The area of moder-ate grade also expanded and decreased alternately from thefirst (2001) to the last year (2015), with the greatest changefrom 2005 to 2010 (about +20%); The eco-environment inthe good and excellent grades expanded continuously from2001 to 2010, then slightly declined to 2015. In general,the extent of poor and excellent grades accounted for less than5% of the study area in four periods, while the area with a fairand moderate eco-environment was dominant.

In order to deeply understand different RSEI gradeschange situation, dynamic transition analysis was performed(Table 3 and Table 4). Generally, JJJ’s eco-environment hadundergone a remarkable change. To be specific, in the firstperiod (2001∼2005), the area of poor and moderate gradedecreased at a rate of 446.94 km2

·y−1 and 2514.38 km2·y−1,

TABLE 2. RSEI grades statistical result.

TABLE 3. RSEI grades dynamic change of JJJ.

respectively; Among it, poor grade area mainly changed intofair grade (4796.50 km2), accounting 99.13% of entire poorgrade changed area; moderate grade areamainly changed intofair grade (19601.75 km2) and good grade (6072.00 km2),accounting 76.34% and 23.65% of entire moderate gradechanged area. Fair, good and excellent grade showed sim-ilar dynamic degree value, which was 2.28%, 2.57% and2.55% per year, respectively. Among it, fair grade areamainlychanged into moderate and poor grade, accounting 79.13%and 19.84% of entire fair grade changed area; as to goodgrade, there had 3414.00 km2 and 458.00 km2 that changedinto the moderate and excellent grade. To excellent grade,the area changing into good grade accounted for 99.77% ofthe total excellent grade changed area.

In the second period (2005∼2010), poor and fair gradedisplayed negative dynamic degree, which was −18.56%and −11.36%, respectively. Besides, moderate, good andexcellent grade showed positive dynamic degree, among

VOLUME 8, 2020 125539

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

TABLE 4. RSEI grades transition matrix of JJJ.

it, the excellent grade had the highest dynamic degree(+29.12%). To be specific, poor grade area mainly changedinto fair grade (4395.75 km2), accounting 98.66% of the totalpoor grade changed area; Fair grade area mainly changedinto moderate grade (66984.25 km2), accounting 99.89%of the total fair grade changed area; Moderate grade areamainly changed into good grade (24049.50 km2), accounting95.76% of the total moderate grade changed area; As togood grade, it mainly changed into excellent and moderategrade, which accounted 78.16% and 21.84% of the total goodgrade changed area, respectively; To excellent grade, therehas 107.75 km2 that changed from excellent grade into goodgrade, accounting 100% of the total excellent grade changedarea.

In the third period (2010∼2015), the poor grade had thehighest dynamic degree (+91.62%), followed by fair grade(+19.35%); however, different from the second period, mod-erate, good and excellent grade all had a negative dynamicdegree, which was −6.26%, −3.78% and −10.05%, respec-tively. Combined with Table 4, poor grade area mainlychanged into fair grade (43.75 km2) and moderate grade(3.25 km2), accounting 93.09% and 6.91% of the total poorgrade changed area; Fair grade mainly changed into poor andmoderate grade, accounting 26.72% and 73.28% of the totalfair grade changed area; Moderate grade mainly convertedto fair grade (51246.00 km2), accounting 95.66% of thetotal moderate grade changed area; To good grade, the areamainly changed into moderate grade (13529.80 km2), whichaccounted 98.91% of the total good grade changed area; As

for the excellent grade, there had 1952.50 km2 that changedinto good grade, accounting 100% of the total excellent gradechanged area.

During the entire period (2001∼2015), poor, fair andmoderate grade all had a negative dynamic degree, whilegood and excellent grade had positive one, representing theeco-environment of JJJ had improved. According to areatransition matrix, poor grade mainly changed into a fairgrade (4820.50 km2) and moderate grade (1027.50 km2),which accounted 82.43% and 17.57% of the total poor gradechanged area, respectively. Different from the former threeperiods, fair grade mainly changed into a moderate grade(25945.05 km2), accounting 93.74% of the total fair gradechanged area. Moderate grade mainly changed into a fairand good grade equally, which accounted for 50.24% and49.68% of the total moderate grade changed area, respec-tively. To good grade, only 6.30% of the total grade area haschanged into other four grades, and mainly were moderategrade (853.25 km2) and excellent grade (706 km2). As for theexcellent grade, it mainly changed into good grade, account-ing 99.70% of the total excellent grade changed area.

In general, in a different period, area transition betweengrades mainly happened in one adjacent grade, which rep-resented that the range of RSEI variation mainly occurredin ±0.2. Moreover, dynamic degree of all grade in the for-mer three periods (2001∼2005, 2005∼2010 and 2010∼2015)demonstrated positive and negative value alternatively. How-ever, in the entire period (2001∼2015), the eco-environmentof JJJ represented an improved trend at the change rate

125540 VOLUME 8, 2020

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

FIGURE 4. Difference results in JJJ during 2001∼2015.

of −329.95 km2· y−1 (poor), −517.55 km2

· y−1 (fair),−225.09 km2

· y−1 (moderate), +1039.73 km2· y−1 (good),

+32.86 km2· y−1 (excellent), respectively. Here, we set

1∼5 to poor, fair, moderate, good and excellent grade, respec-tively; then, difference method was performed by using RSEIgrade in 2015 minus RSEI grade in 2001; finally, we dividedthe image into five types based on the pixel value, whichwere highly degraded {−3,−2}, degraded {−1}, unchanged{0}, improved {1}, and highly improved {2}, respectively(Figure 4).

According to Figure 4, we found that improved regionwas mainly distributed in the northwestern region (Zhangji-akou city). To be specific, these highly improved regionswere mostly located in the central and southwestern partof Zhangjiakou city. This had a great relationship with theseries policies implemented by the government, such as‘Returning Farmland to Forest (grass) Project’, ‘Three-NorthShelter Forest Program’, ‘Beijing-Hebei Ecological WaterResources Protection Forest Project’, etc. The degraded areawas mainly distributed in the eastern Hebei plain, wherethere had intensive anthropogenic activities. One importantreasonwas that construction land and arable landweremainlylocated in this region. To be specific, there had two highlydegraded regions, which were marked as A and B (Figure 4).By importing these regions’ shapefile to Google Earth Pro,we found that Region A was the lake named Angulinao,which was dried since 2004. As for Region B, it belongedto the Beidagang wetland located in Tianjin city. However,

TABLE 5. Four class-level metrics of each RSEI grade in JJJ from 2001 to2015.

in the past years, natural wetlands showed an artificializationtrend, while these artificial wetlands were increasing occu-pied by urban sprawl [70]. Generally, during 2001∼2015,notable effects have been achieved in JJJ by putting a seriesof environmental protection projects into practice. However,more measures should be promoted to seek a coordinateddevelopment between anthropogenic activities and naturalenvironment. Also, we found that these highly degraded andhighly improved regions accounted for a larger proportionof JJJ, indicating that the change of JJJ’s eco-environmentalmost occurred in one RSEI grade.

B. LANDSCAPE PATTERN VARIATIONS OF DIFFERENT RSEIGRADESClass-level landscape metrics can provide addition aggregateproperties at the class level that result from the unique config-uration of patches across the landscape. In this study, after acareful selection, the values of four class-levelmetrics, namedNP, LPI, AREA_MN, and COHESION, are shown in Table 5.Besides, four class-level metrics variations across time wereanalyzed (Figure 5).

NP, a shortened form of ‘number of patches’, repre-sents the number of patches in each RSEI grade. Based onTable 5, the fair and moderate grade had the largest NP value,indicating that they had higher fragmentation. Combinedwith Table 2, in 2005 and 2010, the NP of the fair gradewas 976 and 2047, however, the area was 108418.75 km2

and 46817.75 km2, respectively, indicating that fair gradein 2010 was more fragmented and average patch size wassmall. From 2001 to 2015, only the NP of excellent grade

VOLUME 8, 2020 125541

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

FIGURE 5. Four class-level metrics changes from 2001 to 2015.

increased, further indicating the eco-environment of JJJ hadimproved.

LPI, the abbreviation of ‘largest patch index’, interprets thepercentage of the landscape comprised by the largest patchof each RSEI grade. In 2001, 2005, and 2015, fair grade allacquired the highest LPI value, followed by moderate grade,indicating there existed large clumpy patch of both grades.The LPI values of poor and excellent grades were less than 1,which meant that both grades patches were dispersed.

AREA_MN, shortened of ‘mean patch size’, demonstratedthe average area of all patches in one RSEI grade. Fair,moderate, and good grade all had higher AREA_MN value.From 2001 to 2015, AREA_MN value of poor grade sharplydecreased, excellent grade changed a little, while fair, moder-ate, and good grade all increased partially. Combinedwith NPand LPI, during 2001∼2015, fair, moderate, and good gradeindicated an expansion and more continuous representationof existing dominant patches, however, the decrease of totalfair grade and moderate grade area showed that some smallpatches changed into other grades. Here, the decrease ofNP, the increase of LPI, and the almost unchangeablenessof AREA_MN of excellent grade deeply indicated that theimprovement of JJJ’s eco-environment.

COHESION measures the physical connectedness of thecorresponding patch type. In 2001∼2015, the COHESIONvalues of fair, moderate, and good grades were much high(reaching 100%), representing three grades had extremelygood connectedness. Besides, the COHESION values of poorand excellent grade in 2015 were lower and higher than thatone in 2001, respectively, indicating the decrease of poorgrade area and the increase of excellent grade area.

Generally, combined with Figure 5, from 2001 to 2015,fair, moderate, and good grades were dominant. However,we found that fair grade had the highest NP value, lowest LPI,

TABLE 6. Six landscape-level metrics of each RSEI grade in JJJ from2001 to 2015.

AREA_MN, and COHESION value in 2010, indicating thatfair grade achieved the most fragmented situation. As to mod-erate grade, it had the highest LPI and AREA_MN value, andthe second-highest NP and COHESION value in 2010, rep-resenting that moderate grade accounted for the largest per-centage of JJJ. Compared with 2001, in 2015, fair, moderate,and good grades showed an increase of LPI and AVER_MN,and a decrease of NP, indicating that patches were gettingmore concentrated and continuous; however, the decrease offour metrics of poor grade indicated the more fragmented anddisconnected situation. Besides, the increase of NP, LPI, andCOHESION value of excellent grade conveying the expandedand concentrated situation, more importantly, the improve-ment of JJJ eco-environment. As mentioned above, theacquisition of these changes had a relationship withthe implementation of those environmental improvementprojects, such as ‘Returning Farmland to Forest (grass)Project’, ‘Three-North Shelter Forest Program’, ‘Beijing-Hebei EcologicalWater Resources Protection Forest Project’,etc.

C. LANDSCAPE PATTERN VARIATION OF THE ENTIRE JJJURBAN AGGLOMERATIONLandscape-level metrics can measure the overall structure,function or changes of the entire region by computing allpatches. Besides, these metrics can also be used to inter-pret other characteristics, like fragmentation, connectedness,diversity, etc. In this study, six landscape-level metrics aftercareful selection were applied to identify these features(Table 6).

CONTAG, the abbreviation of ‘contagion index’,is inversely related to edge density. For example, when asingle class occupies a very large percentage of the landscape,the value of CONTAG is high, and vice versa. From 2001 to2015, the value of CONTAG increased first, then decreased,but was still higher than the first period, representing theenhancement of aggregation.

AI, shortened of ‘aggregation index’, measures the level ofaggregation of spatial patterns. From 2001 to 2015, the AIvalue increased by 1.07%, combined with CONTAG, furtherindicating the improvement of patches’ aggregation.

SHDI, the abbreviation of ‘Shannon’s diversity index’, is apopular measure of diversity in community ecology, appliedhere to reflect the diversity of RSEI grades. Compared with2001, SHDI increased to 1.1272 in 2015, representing thepatches were getting more complex.

125542 VOLUME 8, 2020

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

IJI, shortened form of ‘interspersion and juxtapositionindex’, measures the patch adjacency and the degree of theinterspersion or intermixing of patch types. In the fourteenyears, IJI value had increased from 44.2041 to 44.6309, indi-cating the patches of all RSEI grade in JJJ were integratedgradually.

SHAPE_MN, shortened of ‘mean patch shape index’, is theration between the perimeter of a patch and the perimeter ofthe simplest patch in the same area, which can be used toreflect the shape complexity. Compared with 2001, in 2015,the value of this metric decreased from 1.3182 to 1.3015, rep-resenting that the shape of patches was getting more regular.

FRAC_MN, the abbreviation of ‘mean patch fractal dimen-sion’, also measures the patch shape complexity. Sameas SHAPE_MN metric, the value of FRAC_MN metricdecreased from 1.0316 to 1.0306, further demonstrating thatpatches’ shape was more regular.

In general, based on six metrics in landscape level,we found that the patches of all RSEI grades in JJJ weregetting more aggregated, connected, diverse, and regular.

D. RSEI CHANGE IMPACT ON LANDSCAPE PATTERNThe changes of RSEI grades have a direct influence on land-scape pattern. The above content has analyzed the RSEI gradearea change situation and the landscape pattern variations inboth class-level and landscape-level. Here, the relationshipbetween RSEI grades and landscape-level metrics was ana-lyzed (Figure 6). The order of six metrics average R2 wasCONTAG (R2

= 0.0263) < AI (R2= 0.1397) < SHDI

(R2= 0.2796) < IJI (R2

= 0.5612) < SHAPE_MN (R2=

0.7921) < FRAC_MN (R2= 0.8898). CONTAG metric had

an extremely weak correlation with five RSEI grades, withthe average R2 of 0.0263, representing that the changes ofRSEI grade area hardly exerted influence on the variation ofCONTAG metric; As to AI metric, it had a higher correlationwith fair and moderate grades, which were the two dominantgrades. The left four metrics (SHDI, IJI, SHAPE_MN, andFRAC_MN) all had a stronger correlation with fair and mod-erate grades, moreover, poor, good, and excellent grades allhad an increasing correlation.

Generally, except those poor correlations, five RSEI gradesall had relatively good correlation, representing the changesof RSEI grades influenced the variation of the landscape.However, the effects were different. Poor and Fair gradeshad a positive correlation with landscape-level metrics, whilemoderate, good, and excellent grades had a negative one.Besides, taking the small area proportion occupied by poorand excellent grades (total less than 5%) into consideration,the changes of left three grades showed a more dominant andstronger influence on the variation of landscape pattern in JJJ.

IV. DISCUSSIONA. COMPARISON OF ECO-ENVIRONMENT QUALITY FROMRSEI AND EIIn our study, ecological index (EI), announced by theministryof ecology and environment of China [71], was adopted to

compare its result with the RSEI result. Here, to compareEI and RSEI at the same dimension, RSEI value was mul-tiplied by 100, accordingly, the range of five grades wasalso multiplied by 100. Besides, the grading standard of EIwas also given by the ecological index technical criterion,which was poor [0, 20), fair [20, 35), moderate [35, 55), good[55, 75), and excellent [75, 100), respectively. Table 7 wasa comparison of thirteen cities’ EI and RSEI calculationresults. Among it, the changing trend column had two types,the upward-pointing arrow meant the eco-environment ofthis city during the fourteen years had improved, while thedownward-pointing arrow meant the eco-environment of thiscity had deteriorated. Table 8 was a comparison of EI andRSEI in JJJ during 2001∼2015.

According to Table 7, we found that EI value ofeach city showed different change trend. To be specific,from 2001 to 2015, Beijing, Tianjin, Shijiazhuang, Tang-shan, Handan, Baoding, Zhangjiakou, Chengde, Cangzhou,and Langfang all showed an increasing trend; however,Qinhuangdao, Xingtai, and Hengshui all displayed adecreased trend. As for RSEI, Beijing, Qinhuangdao, Baod-ing, Zhangjiakou, Chengde, Cangzhou, and Shijiazhuang alldisplayed an increasing trend, while Tianjin, Tangshan, Han-dan, Xingtai, Langfang, and Hengshui showed the oppositeone. Generally, eight cities showed the same change trend inboth EI and RSEI, five cities showed the opposite changetrend. This might have a relationship with the differencebetween the two indexes’ data acquisition and calculationmethods. The RSEI was purely calculated and derived fromremote sensing datasets, therefore, it could reflect the region’seco-environment anywhere. As for EI, it not only relied onpart remote sensing data but also relied on statistical data.More importantly, statistical data could not be acquired datathe gridded level, therefore using one value to represent theentire region might be inappropriate, which might have led totheir difference. Based on the grading standard of each index,we compare them in four periods (Table 8). The grades of EIin 2001, 2005, and 2010 all belonged to fair grade, while thegrades of RSEI in the same years belonged tomoderate grade.In 2015, RSEI and EI all belonged to the moderate grade.In general, RSEI and EI belonged to different grade in 2001,2005, and 2010, but the EI value in three years was close to thethreshold value (35); moreover, the RSEI value in 2001 and2015 was just slightly higher than the threshold value (40),which meant that there existed little difference between RSEIand EI. More importantly, during 2001∼2015, RSEI and EIrepresented the same conclusion that the eco-environmentof JJJ had improved, which further validated our findings.Numerous previous studies have validated the RSEI indexfrom several aspects at the city or provincial level. For exam-ple, Yue et al. [38] constructed one PSR evaluating indexsystem, which integrated remote sensing data, statistical data,meteorological data, and so forth, to compare its results withRSEI result, although the area percentage of five grades wasdifferent, the spatial distribution of their results was quitesimilar with each other; Xu et al. [36] also found that RSEI

VOLUME 8, 2020 125543

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

FIGURE 6. Relationships between six landscape-level metrics changes and the area percentage of RSEI poor, fair, moderate, good, and excellent grades.(a1) ∼(a5) were five RSEI grades and CONTAG. (b1) ∼(b5) were five RSEI grades and AI. (c1) ∼(c5) were five RSEI grades and SHDI. (d1) ∼(d5) were fiveRSEI grades and IJI. (e1) ∼(e5) were five RSEI grades and SHAPE_MN. (f1) ∼(f5) were five RSEI grades and FRAC_MN.

125544 VOLUME 8, 2020

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

TABLE 7. A comparison of EI and RSEI calculation results of four years.

TABLE 8. A comparison of average EI and RSEI during 2001∼2015.

showed a small difference in monitoring eco-environmentcondition by comparing EI and RSEI results. Generally,although there existed a small difference, RSEI was provento be an effective index to evaluate region eco-environment,whether in the city level, provincial level or large level, likean urban agglomeration.

B. THE DRIVING FORCING OF JJJ’S ECO-ENVIRONMENTCHANGETo explore the driving factors’ influence on JJJ’s eco-environment change in a different year, we have carefullyselected three natural factors, which were annual averagetemperature (AT), annual precipitation (PR), and eleva-tion (EV), and three human factors, which were land use(LU), gross regional product per square kilometre (GP), andpopulation density (PD) [72], [73]. The Geodetertor methodwas one method that could investigate the spatially strati-fied heterogeneity of the geographic variable Y and explorehow factor X explains the spatial pattern of Y [74], [75].The importance of a factor could be represented by the qvalue, which ranges from 0 to 1. A higher q value indicatesthat Y has a stronger spatially stratified heterogeneity andfactor X can explain 100 × q of the spatial pattern of Y .In this study, Y represents the RSEI gridded result in 2001,2005, 2010, and 2015; X represents each driving factor.Table 9 was six driving factors q values in 2001, 2005, 2010,and 2015.

TABLE 9. Six driving factors q values in 2001, 2005, 2010, and 2015.

According to Table 9, we found that, in the third natu-ral factors, AT and EV had an increased influence during2001∼2015, however, PR had a decreased one. Among it,AT and EV could explain a relatively higher spatial stratifiedheterogeneity of the RSEI, showing that the spatial distri-bution of AT and EV might be consistent with the RSEIspatial distribution; as for PR, the lower q value representedthat it had limited influence on the JJJ’s eco-environmentchanges, In the third human factors, LU, GP, and PD allrepresented an increased influence from 2001 to 2015. To bespecific, three human factors had the highest q value in 2015,which reflected that the intensity of anthropogenic activitieswas also getting stronger, which might be connected withthe urbanization process [7]. Generally, the eco-environmentof JJJ in the different years was driven by both naturaland human factors, and during the process of urbanization,anthropogenic activities could play an important role inchanging the regional eco-environment [76], [77].

C. IMPLICATIONS FOR REGIONAL ECOLOGICALLANDSCAPE MANAGEMENTThe analysis of investigating the relationship betweenlandscape-level metrics and RSEI grades percentage rep-resented that mostly landscape-level metrics had a pos-itive or negative correlation. Although in the field of

VOLUME 8, 2020 125545

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

region heat islands or LULC studies, the relationship wasexplored [47], [78], our finding was still interesting. As ourfinding firstly validated that there existed a relationshipbetween different RSEI grades and landscape-level metrics.Moreover, it provided some implications for regional ecolog-ical landscape management. As a better spatial configurationcould be achieved based on these relationships. To be specific,if the regional ecological landscape managers want the wholeregion’s landscape pattern more regular, then based on therelationship, they can try to decrease the area of poor andfair grades, and increase the area of moderate, good, andexcellent grades. Considering the area percentage of differentgrades, focusing on the fair, moderate, and good three dom-inant grades can achieve the regular landscape pattern moreefficiently. Generally, our finding may provide new insightfor the regional eco-environment landscape management andconfiguration.

D. LIMITATIONS AND FURTHER STUDYEven the variations of JJJ’s eco-environment and the dynam-ics of landscape pattern have been investigated, there stillexist some limitations. Firstly, only four-time periods havebeen performed, however, more deep information needs tobe studied. For instance, whether the eco-environment existsnatural fluctuations in the scale of per year still needs to dis-cover. Secondly, we only selected limited landscape patternmetrics to study the relationship, whether these metrics canappropriately reveal these relationships still needs to explorein the future. Thirdly, more driving factors influencing theeco-environment also needs to investigate. Finally, simulationtools may help achieve the optimal eco-environment land-scape pattern and providing useful suggestions for relevantpolicy-makers.

V. CONCLUSIONSFour remote sensing ecological index (RSEI) maps of JJJfrom 2001 to 2015 were firstly acquired and equally divided.Then spatiotemporal changes of RSEI and landscape patternvariation in class-level and landscape-level were analyzed.Finally, the relationship between the area percentage of fivegrades and six landscape-level metrics were performed.

This study found that the eco-environment of JJJ hadimproved in the fourteen years, with the RSEI valueincreased from 0.43 to 0.46, with the improved region wasmainly located in Zhangjiakou city, while the degradedregion was mainly distributed in the eastern Hebei plain.We also found that the landscape characteristics of entireJJJ eco-environment were becoming more aggregated, con-nected, diverse, and regular. Besides, this study found thatthe eco-environment of JJJ was dominated by three RSEIgrades, which were fair, moderate, and good. More impor-tantly, we found that anthropogenic activities exert increas-ingly importance on the regional eco-environment changesand different RSEI grades had a positive or negative cor-relation with the variation of landscape pattern, which mayprovide an view in managing and achieving the optimal eco-environment landscape pattern.

REFERENCES[1] J. Song, H. X. Wang, and F. Wang, ‘‘Research progress of ecological envi-

ronment quality assessment and methods review,’’ (in Chinese), Environ.Sci. Technol., vol. 36, no. S2, pp. 448–453, 2013.

[2] Z.-W. Li, G.-M. Zeng, H. Zhang, B. Yang, and S. Jiao, ‘‘The integratedeco-environment assessment of the red soil hilly region based on GIS—A case study in Changsha City, China,’’ Ecol. Model., vol. 202, nos. 3–4,pp. 540–546, Apr. 2007.

[3] Q. Wen, Z. Zhang, L. Shi, X. Zhao, F. Liu, J. Xu, L. Yi, B. Liu, X. Wang,L. Zuo, S. Hu, N. Li, and M. Li, ‘‘Extraction of basic trends of urbanexpansion in China over past 40 years from satellite images,’’ Chin. Geo-graphical Sci., vol. 26, no. 2, pp. 129–142, Apr. 2016.

[4] L. Shi, H. Taubenböck, Z. Zhang, F. Liu, and M. Wurm, ‘‘Urbanization inChina from the end of 1980s until 2010—Spatial dynamics and patternsof growth using EO-data,’’ Int. J. Digit. Earth, vol. 12, no. 1, pp. 78–94,Jan. 2019.

[5] G. Zhao, X. Mu, Z. Wen, F. Wang, and P. Gao, ‘‘Soil erosion conserva-tion, and eco-environment changes in the Loess Plateau of China,’’ LandDegradation Develop., vol. 24, no. 5, pp. 499–510, Sep. 2013.

[6] R. C. Estoque, Y. Murayama, and S. W. Myint, ‘‘Effects of landscapecomposition and pattern on land surface temperature: An urban heat islandstudy in the megacities of southeast Asia,’’ Sci. Total Environ., vol. 577,pp. 349–359, Jan. 2017.

[7] J. Haas and Y. Ban, ‘‘Urban growth and environmental impacts inJing-Jin-Ji, theYangtze, River Delta and the Pearl River Delta,’’ Int. J. Appl.Earth Observ. Geoinf., vol. 30, pp. 42–55, Aug. 2014.

[8] S. Dey, L. Di Girolamo, A. van Donkelaar, S. N. Tripathi, T. Gupta, andM. Mohan, ‘‘Variability of outdoor fine particulate (PM2.5) concentrationin the Indian subcontinent: A remote sensing approach,’’ Remote Sens.Environ., vol. 127, pp. 153–161, Dec. 2012.

[9] M. Mukherjee, A. K. Ray, and C. Rajyalakshmi, ‘‘Physical quality of lifeindex: Some international and Indian applications,’’ Social Indicators Res.,vol. 6, no. 3, pp. 283–292, Jul. 1979.

[10] I. van Kamp, K. Leidelmeijer, G. Marsman, and A. de Hollander, ‘‘Urbanenvironmental quality and humanwell-being: Towards a conceptual frame-work and demarcation of concepts; a literature study,’’ Landsc. Urban.Plan., vol. 65, nos. 1–2, pp. 5–18, 2003.

[11] B. J. Fu, ‘‘The evaluation of eco-environmental qualities in China,’’ (inChinese), China Popul. Resour. Environ., vol. 2, no. 2, pp. 48–54, 1992.

[12] N. B. Grimm, S. H. Faeth, N. E. Golubiewski, C. L. Redman, J. G. Wu,X. M. Bai, and J. M. Briggs, ‘‘Global change and the ecology of cities,’’Science, vol. 319, no. 5864, pp. 756–760, 2008.

[13] L. L. Lu, Q. H. Weng, H. D. Guo, S. Y. Feng, and Q. T. Li, ‘‘Assessment ofurban environmental change using multi-source remote sensing time series(2000–2016): A comparative analysis in selected megacities in Eurasia,’’Sci. Total Environ., vol. 684, pp. 553–564, Sep. 2019.

[14] Z. Y. Ouyang, X. K. Wang, and H. Miao, ‘‘China’s eco-environmentalsensitivity and its spatial heterogeneity,’’ (in Chinese), Acta Ecol. Sinica,vol. 20, no. 1, pp. 9–12, 2000.

[15] X. Dai, Z. Li, S. Lin, and W. Xu, ‘‘Assessment and zoning of eco-environmental sensitivity for a typical developing province in China,’’Stochastic Environ. Res. Risk Assessment, vol. 26, no. 8, pp. 1095–1107,Dec. 2012.

[16] M. Su, H. Xie, W. Yue, L. Zhang, Z. Yang, and S. Chen, ‘‘Urban ecosystemhealth evaluation for typical Chinese cities along the belt and road,’’ Ecol.Indicators, vol. 101, pp. 572–582, Jun. 2019.

[17] C. D. Yan, H. J. Dai, andW. Guo, ‘‘Evaluation of ecological environmentalquality in a coal mining area by modelling approach,’’ Sustainability.,vol. 9, no. 8, pp. 1265–1277, 2017.

[18] R. Li, ‘‘Dynamic assessment on regional eco-environmental quality usingAHP-statistics model—A case study of Chaohu lake basin,’’ Chin. Geo-graphical Sci., vol. 17, no. 4, pp. 341–348, Dec. 2007.

[19] C.-L. Miao, L.-Y. Sun, and L. Yang, ‘‘The studies of ecological environ-mental quality assessment in Anhui province based on ecological foot-print,’’ Ecol. Indicators, vol. 60, pp. 879–883, Jan. 2016.

[20] K. Sinha and P. D. Saha, ‘‘Assessment of water quality index using clus-ter analysis and artificial neural network modeling: A case study of theHooghly river basin, West Bengal, India,’’ Desalination Water Treatment,vol. 54, no. 1, pp. 28–36, Apr. 2015.

[21] Z. Y. Li and X. M. Deng, ‘‘Two classes assessment of comprehensiveenvironmental quality based on B-P neural network,’’ (in Chinese), Res.Environ. Sci., vol. 8, no. 3, pp. 32–35, 1995.

125546 VOLUME 8, 2020

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

[22] Z. Y. Li, ‘‘Assessment on the comprehensive urban environmental qualitybased on a matter element analysis,’’ (in Chinese), Chin. J. Environ.,vol. 16, no. 5, pp. 76–78, 1995.

[23] H. Y. Wu, K. L. Chen, Z. H. Chen, Q. H. Chen, Y. P. Qiu, J. C. Wu, andJ. F. Zhang, ‘‘Evaluation for the ecological quality status of coastal watersin east China sea using fuzzy integrated assessment method,’’Mar. Pollut.Bull., vol. 64, no. 3, pp. 546–555, Mar. 2012.

[24] Y. Xu and H. R. Zhou, ‘‘A preliminary study on advances in assessmentof eco-environmental quality in China,’’ (in Chinese), Arid Land Geogr.,vol. 26, no. 2, pp. 166–172, 2003.

[25] J. W. Rouse, R. H. Haas, J. A. Schell, and D. W. Deering, ‘‘Monitoringvegetation systems in the great plains with ERTS,’’ in Proc. Int. Conf. 3rdERTS Symp., Washington, DC, USA, 1973, pp. 309–317.

[26] D. Alcaraz-Segura, A. Lomba, R. Sousa-Silva, D. Nieto-Lugilde, P. Alves,D. Georges, J. R. Vicente, and J. P. Honrado, ‘‘Potential of satellite-derivedecosystem functional attributes to anticipate species range shifts,’’ Int.J. Appl. Earth Observ. Geoinf., vol. 57, pp. 86–92, May 2017.

[27] Z. Yu, Y. Yao, G. Yang, X. Wang, and H. Vejre, ‘‘Spatiotemporal patternsand characteristics of remotely sensed region heat islands during the rapidurbanization (1995–2015) of southern China,’’ Sci. Total Environ., vol. 674,pp. 242–254, Jul. 2019.

[28] C. He, B. Gao, Q. Huang, Q. Ma, and Y. Dou, ‘‘Environmental degradationin the urban areas of China: Evidence from multi-source remote sensingdata,’’ Remote Sens. Environ., vol. 193, pp. 65–75, May 2017.

[29] W. Wei, Z. Guo, B. Xie, J. Zhou, and C. Li, ‘‘Spatiotemporal evolutionof environment based on integrated remote sensing indexes in arid inlandriver basin in northwest China,’’ Environ. Sci. Pollut. Res., vol. 26, no. 13,pp. 13062–13084, May 2019.

[30] W. Wei, S. Shi, X. Zhang, L. Zhou, B. Xie, J. Zhou, and C. Li, ‘‘Regional-scale assessment of environmental vulnerability in an arid inland basin,’’Ecol. Indicators, vol. 109, Feb. 2020, Art. no. 105792.

[31] Y. Chang, K. Hou, Y. Wu, X. Li, and J. Zhang, ‘‘A conceptual frameworkfor establishing the index system of ecological environment evaluation—A case study of the upper Hanjiang river, China,’’ Ecol. Indicators,vol. 107, Dec. 2019, Art. no. 105568.

[32] L. H. Chai and D. Lha, ‘‘A new approach of deriving indicators andcomprehensive measure for ecological environmental quality assessment,’’Ecol. Indicators, vol. 85, pp. 716–728, Feb. 2018.

[33] R. Sun, Z.Wu, B. Chen, C. Yang, D. Qi, G. Lan, and K. Fraedrich, ‘‘Effectsof land-use change on eco-environmental quality in Hainan island, China,’’Ecol. Indicators, vol. 109, Feb. 2020, Art. no. 105777.

[34] M. C. Yan and Y. C. Wang, ‘‘Advances in the evaluation of ecologicalenvironmental quality,’’ (in Chinese), Ecol. Environ. Sci., vol. 21, no. 10,pp. 1781–1788, 2012.

[35] H. Q. Xu, ‘‘A remote sensing urban ecological index and its application,’’(in Chinese), Acta Ecol. Sinica, vol. 33, no. 24, pp. 7853–7862, 2013.

[36] H. Q. Xu, Y. F. Wang, H. D. Guan, T. T. Shi, and X. S. Hu, ‘‘Detectingecological changes with a remote sensing based ecological index (RSEI)produced time series and change vector analysis,’’ Remote Sens., vol. 11,pp. 2345–2371, 2019.

[37] H. Kang,W. Tao, Y. Chang, Y. Zhang, L. Xuxiang, and P. Chen, ‘‘A feasiblemethod for the division of ecological vulnerability and its driving forces insouthern Shaanxi,’’ J. Cleaner Prod., vol. 205, pp. 619–628, Dec. 2018.

[38] H. Yue, Y. Liu, Y. Li, and Y. Lu, ‘‘Eco-environmental quality assessment inChina’s 35 major cities based on remote sensing ecological index,’’ IEEEAccess, vol. 7, pp. 51295–51311, 2019.

[39] X. Hu and H. Xu, ‘‘A new remote sensing index for assessing the spa-tial heterogeneity in urban ecological quality: A case from Fuzhou city,China,’’ Ecol. Indicators, vol. 89, pp. 11–21, Jun. 2018.

[40] X. Hu and H. Xu, ‘‘A new remote sensing index based on the pressure-state-response framework to assess regional ecological change,’’ Environ.Sci. Pollut. Res., vol. 26, no. 6, pp. 5381–5393, Feb. 2019.

[41] H. Xu, M. Wang, T. Shi, H. Guan, C. Fang, and Z. Lin, ‘‘Prediction ofecological effects of potential population and impervious surface increasesusing a remote sensing based ecological index (RSEI),’’ Ecol. Indicators,vol. 93, pp. 730–740, Oct. 2018.

[42] P.-G. Cheng, C.-Z. Tong, X.-Y. Chen, and Y.-J. Nie, ‘‘Urban ecologicalenvironment monitoring and evaluation based on remote sensing ecologi-cal index,’’ in Proc. Int. Conf. Intell. Earth Observing Appl., Guilin, China,Dec. 2015, p. 98083.

[43] W. Shan, X. Jin, J. Ren, Y. Wang, Z. Xu, Y. Fan, Z. Gu, C. Hong, J. Lin,and Y. Zhou, ‘‘Ecological environment quality assessment based on remotesensing data for land consolidation,’’ J. Cleaner Prod., vol. 239, Dec. 2019,Art. no. 118126.

[44] H. L. Guo, B. W. Zhang, Y. F. Bai, and X. H. He, ‘‘Ecological environmentassessment based on remote sensing in Zhengzhou,’’ in Proc. Int. Conf.Energy Environ. Mater. Sci., Singapore, 2017, Art. no. 012190.

[45] D. Meng, S. Yang, H. Gong, X. Li, and J. Zhang, ‘‘Assessment of thermalenvironment landscape over five megacities in China based on Landsat 8,’’J. Appl. Remote Sens., vol. 10, no. 2, Jun. 2016, Art. no. 026034.

[46] J. Wu, ‘‘Landscape ecology, cross-disciplinarity, and sustainability sci-ence,’’ Landscape Ecol., vol. 21, no. 1, pp. 1–4, Jan. 2006.

[47] L. Wang, S. Wang, Y. Zhou, J. Zhu, J. Zhang, Y. Hou, and W. Liu,‘‘Landscape pattern variation, protection measures, and land use/landcover changes in drinking water source protection areas: A case studyin Danjiangkou Reservoir, China,’’ Global Ecol. Conservation, vol. 21,Mar. 2020, Art. no. e00827.

[48] S. A. Cushman, K. McGarigal, and M. C. Neel, ‘‘Parsimony in landscapemetrics: Strength, universality, and consistency,’’ Ecol. Indicators, vol. 8,no. 5, pp. 691–703, Sep. 2008.

[49] C. Fan and S. Myint, ‘‘A comparison of spatial autocorrelationindices and landscape metrics in measuring urban landscapefragmentation,’’ Landscape Urban Planning, vol. 121, pp. 117–128,Jan. 2014.

[50] R. Hao, D. Yu, Y. Liu, Y. Liu, J. Qiao, X. Wang, and J. Du, ‘‘Impacts ofchanges in climate and landscape pattern on ecosystem services,’’ Sci. TotalEnviron., vol. 579, pp. 718–728, Feb. 2017.

[51] S. N. Gillanders, N. C. Coops, M. A. Wulder, S. E. Gergel, and T. Nelson,‘‘Multitemporal remote sensing of landscape dynamics and pattern change:Describing natural and anthropogenic trends,’’ Prog. Phys. Geography,Earth Environ., vol. 32, no. 5, pp. 503–528, Oct. 2008.

[52] C. G. K. Boongaling, D. V. Faustino-Eslava, and F. P. Lansigan, ‘‘Mod-eling land use change impacts on hydrology and the use of landscapemetrics as tools for watershed management: The case of an ungaugedcatchment in the Philippines,’’ Land Use Policy, vol. 72, pp. 116–128,Mar. 2018.

[53] A. Lausch, T. Blaschke, D. Haase, F. Herzog, R.-U. Syrbe, L. Tischen-dorf, and U. Walz, ‘‘Understanding and quantifying landscape structure—A review on relevant process characteristics, data models and landscapemetrics,’’ Ecol. Model., vol. 295, pp. 31–41, Jan. 2015.

[54] C. Echeverría, A. Newton, L. Nahuelhual, D. Coomes, andJ. M. Rey-Benayas, ‘‘How landscapes change: Integration of spatialpatterns and human processes in temperate landscapes of southern Chile,’’Appl. Geography, vol. 32, no. 2, pp. 822–831, Mar. 2012.

[55] Y. Chi, Z. Zhang, J. Gao, Z. Xie, M. Zhao, and E. Wang, ‘‘Evaluatinglandscape ecological sensitivity of an estuarine island based on landscapepattern across temporal and spatial scales,’’ Ecol. Indicators, vol. 101,pp. 221–237, Jun. 2019.

[56] L. S. Bertolo, G. T. N. P. Lima, and R. F. Santos, ‘‘Identifying changetrajectories and evolutive phases on coastal landscapes. Case study: SãoSebastião Island, Brazil,’’ Landscape Urban Planning, vol. 106, no. 1,pp. 115–123, May 2012.

[57] H. Dadashpoor, P. Azizi, and M. Moghadasi, ‘‘Land use change, urbaniza-tion, and change in landscape pattern in a metropolitan area,’’ Sci. TotalEnviron., vol. 655, pp. 707–719, Mar. 2019.

[58] A. I. R. Cabral and F. L. Costa, ‘‘Land cover changes and landscape patterndynamics in Senegal and Guinea Bissau borderland,’’ Appl. Geography,vol. 82, pp. 115–128, May 2017.

[59] F. Kienast, J. Frick, M. J. van Strien, and M. Hunziker, ‘‘The swisslandscapemonitoring program—Acomprehensive indicator set tomeasurelandscape change,’’ Ecol. Model., vol. 295, pp. 136–150, Jan. 2015.

[60] A. Chen, L. Yao, R. Sun, and L. Chen, ‘‘How many metrics are required toidentify the effects of the landscape pattern on land surface temperature?’’Ecol. Indicators, vol. 45, pp. 424–433, Oct. 2014.

[61] C. Fang and D. Yu, ‘‘Urban agglomeration: An evolving conceptof an emerging phenomenon,’’ Landscape Urban Planning, vol. 162,pp. 126–136, Jun. 2017.

[62] S. Yang and L. Shi, ‘‘Prediction of long-term energy consumption trendsunder the new national urbanization plan in China,’’ J. Cleaner Prod.,vol. 166, pp. 1144–1153, Nov. 2017.

[63] Z. M. Wan, ‘‘New refinements and validation of the collection-6 MODISland-surface temperature/emissivity product,’’ Remote Sens. Environ.,vol. 140, pp. 136–145, Jan. 2014.

[64] X. Zhang, C. B. Schaaf, M. A. Friedl, A. H. Strahler, F. Gao, andJ. C. F. Hodges, ‘‘MODIS tasseled cap transformation and its utility,’’in Proc. IEEE Int. Geosci. Remote Sens. Symp., Toronto, ON, Canada,Jun. 2002, pp. 1063–1065.

VOLUME 8, 2020 125547

J. Ji et al.: Spatiotemporal Change and Landscape Pattern Variation of Eco-Environmental Quality

[65] Z. Yu, X. Guo, Y. Zeng, M. Koga, and H. Vejre, ‘‘Variations in land surfacetemperature and cooling efficiency of green space in rapid urbanization:The case of Fuzhou city, China,’’Urban Forestry Urban Greening, vol. 29,pp. 113–121, Jan. 2018.

[66] K. McGarigal, S. A. Cushman, and E. Ene, ‘‘FRAGSTATS v4:Spatial pattern analysis program for categorical and continuousmaps,’’ Comput. Softw. Program, Univ. Massachusetts, Amherst,Amherst, MA, USA, Tech. Rep., 2012. [Online]. Available:http://www.umass.edu/landeco/research/fragstats/fragstats.html

[67] S. G. Plexida, A. I. Sfougaris, I. P. Ispikoudis, and V. P. Papanastasis,‘‘Selecting landscape metrics as indicators of spatial heterogeneity—A comparison among Greek landscapes,’’ Int. J. Appl. Earth Observ.Geoinf., vol. 26, pp. 26–35, Feb. 2014.

[68] F. Aguilera, L. M. Valenzuela, and A. Botequilha-Leitão, ‘‘Landscapemetrics in the analysis of urban land use patterns: A case study in aSpanish metropolitan area,’’ Landscape Urban Planning, vol. 99, nos. 3–4,pp. 226–238, Mar. 2011.

[69] E. Uuemaa, Ü. Mander, and R. Marja, ‘‘Trends in the use of landscape spa-tial metrics as landscape indicators: A review,’’ Ecol. Indicators, vol. 28,pp. 100–106, May 2013.

[70] H. Y. Huo, J. F. Guo, Z. L. Li, and X. G. Jiang, ‘‘Remote sensing ofspatiotemporal changes in wetland geomorphology based on type 2 fuzzysets: A case study of Beidagangwetland from 1975 to 2015,’’Remote Sens.,vol. 9, no. 7, pp. 683–706, 2017.

[71] Technical Criterion for Ecosystem Status Evaluation, documentHJ 192-2015, 2015.

[72] Y. Li and G. Liu, ‘‘Characterizing spatiotemporal pattern of land usechange and its driving force based on GIS and landscape analysis tech-niques in Tianjin during 2000–2015,’’ Sustainability, vol. 9, no. 6, p. 894,May 2017.

[73] J. He, Z. Pan, D. Liu, and X. Guo, ‘‘Exploring the regional differencesof ecosystem health and its driving factors in China,’’ Sci. Total Environ.,vol. 673, pp. 553–564, Jul. 2019.

[74] T. Shi, Z. Hu, Z. Shi, L. Guo, Y. Chen, Q. Li, and G.Wu, ‘‘Geo-detection offactors controlling spatial patterns of heavy metals in urban topsoil usingmulti-source data,’’ Sci. Total Environ., vol. 643, pp. 451–459, Dec. 2018.

[75] J. Wang, X. Li, G. Christakos, Y. Liao, T. Zhang, X. Gu, and X. Zheng,‘‘Geographical detectors-based health risk assessment and its applicationin the neural tube defects study of the Heshun region, China,’’ Int. J. Geo-graphical Inf. Sci., vol. 24, no. 1, pp. 107–127, Jan. 2010.

[76] S. Wang, H. Ma, and Y. Zhao, ‘‘Exploring the relationship between urban-ization and the eco-environment—A case study of Beijing-Tianjin-Hebeiregion,’’ Ecol. Indicators, vol. 45, pp. 171–183, Oct. 2014.

[77] Z. Wang, L. Liang, Z. Sun, and X. Wang, ‘‘Spatiotemporal differentia-tion and the factors influencing urbanization and ecological environmentsynergistic effects within the Beijing-Tianjin-Hebei urban agglomeration,’’J. Environ. Manage., vol. 243, pp. 227–239, Aug. 2019.

[78] Z. Yu, Y. Yao, G. Yang, X. Wang, and H. Vejre, ‘‘Strong contributionof rapid urbanization and urban agglomeration development to regionalthermal environment dynamics and evolution,’’ Forest Ecol. Manage.,vol. 446, pp. 214–225, Aug. 2019.

JIANWAN JI received the M.S. degree in car-tography and geography information system fromFujian Normal University, Fuzhou, China, in 2018.He is currently pursuing the Ph.D. degree incartography and geography information systemfrom the Aerospace Information Research Insti-tute, Chinese Academy of Sciences. His researchinterest includes the application of remote sensingin the human living environment.

SHIXIN WANG is currently a Professor and Ph.D.Supervisor in cartography and geography informa-tion system at theAerospace Information ResearchInstitute, Chinese Academy of Sciences. His cur-rent research interests include disaster and envi-ronmental remote sensing.

YI ZHOU is currently a Professor and Ph.D. Super-visor in cartography and geography informationsystem at the Aerospace Information ResearchInstitute, Chinese Academy of Sciences. Her cur-rent research interests include disaster and envi-ronmental remote sensing.

WENLIANG LIU received the Ph.D. degree incartography and geography information systemfrom the Institute of Remote SensingApplications,Chinese Academy of Sciences, Beijing, China,in 2011. He is currently an Associate Profes-sor at the Aerospace Information Research Insti-tute, Chinese Academy of Sciences. His currentresearch interests include disaster and environ-mental remote sensing.

LITAO WANG received the Ph.D. degree in car-tography and geography information system fromthe Institute of Remote Sensing Applications,Chinese Academy of Sciences, Beijing, China,in 2007. He is currently an Associate Profes-sor at the Aerospace Information Research Insti-tute, Chinese Academy of Sciences. His currentresearch interests include disaster and environ-mental remote sensing.

125548 VOLUME 8, 2020